Analysis of large-scale climate conditions associated with extreme river flow is an important first step in the development of predictive relationships for such events. The potential of this approach is demonstrated here for the Waitaki River (a river of national importance in terms of electricity generation), in the Southern Alps of New Zealand. Here, atmospheric circulation anomalies and air parcel trajectories associated with such events are investigated for the period 1960–2010, using the NCEP/NCAR reanalysis and HYSPLIT trajectory model. Results show that atmospheric circulation variation and air parcel trajectories associated with extreme high Waitaki river flow events typically follow two distinct patterns. These patterns are associated with differences in both New Zealand- and hemispheric-scale atmospheric circulation, but all occur under a similar pattern of monthly average pressure anomalies. As such, the results indicate that different precipitation generation mechanisms are captured by a single monthly climate anomaly pattern – providing substantial new understanding of the cascade of processes linking atmospheric to surface hydrological variation in the Southern Alps, and pointing the direction for future process-informed research on sources of predictability for Waitaki river flow.
Analysis of the immediate large-scale climate drivers of extreme high river flow is an important first step in the development of predictive relationships for such events. Prediction of river flow is a particularly important research goal from an applied perspective when river water is used for electricity generation, as is the case for the Waitaki River in New Zealand (the seven power stations on this river comprise 17 % of national electricity generation; Electricity Authority, 2011). As well as causing potentially damaging floods, extreme river flow is challenging from a management perspective, and potentially costly (water spilled from dams cannot be used for electricity generation).
The position of New Zealand in the mid-latitude of the southern hemisphere
results in a strong link between the climate of New Zealand and the general
southern hemisphere atmospheric circulation. In particular, the climate of
New Zealand is influenced by both the Southern Annular Mode and the Southern
Oscillation (Gordon, 1986; Kidston et al., 2009). At the same time, the
presence of a substantial topographic barrier (the Southern Alps; peaks
It has been demonstrated previously that weak-to-moderate relationships
exist between the SO and Waitaki river flow on a lagged seasonal basis
(e.g. McKerchar et al., 1998). Predictive relationships of similar strength for
the Waitaki have also been found between geopotential height and SST at a
series of locations near to and far from New Zealand (Purdie and Bardsley,
2010). In terms of concurrent atmosphere-river flow relationships, it has
been shown that a series of circulation indices that describe the meridional
and zonal components of atmospheric circulation over New Zealand (and along
the axis of the Southern Alps in particular: the MZ1 and MZ2) have a strong
relationship to Pukaki inflow (correlation coefficients
Despite the occurrence of atmosphere-river flow relationships at the monthly timescale, the atmospheric processes that cause individual extreme events can only be inferred from statistical relationships at this temporal resolution. Indeed, such time-averaged relationships may mask some atmospheric variation associated with extremes of discharge, thus limiting their usefulness as a basis for understanding the causes (and so predictability) of such events. In response to this research gap, here the applicability of previously identified relationships at the monthly timescale (Kingston et al., 2014) to daily and event-scale flow extremes is investigated for the Waitaki river, focussing specifically on inflow to Lake Pukaki (an important headwater lake of the Waitaki). There are two steps to this analysis: firstly, investigation of daily geopotential height fields associated with extreme high river flow events, and secondly, of precipitation delivery mechanisms for specific extreme flow events (identified using back-trajectory analysis). The situation of the trajectories within the daily circulation anomalies is also analysed, together with a comparison against previously described monthly-resolution relationships (Kingston et al., 2014). The resultant improved understanding of the cascade of processes linking atmospheric to extreme surface hydrological variation will ultimately inform further development of lag-lead relationships for prediction of extremes and river flow more generally.
The Waitaki river (11 900 km
Climate within the upper Waitaki basin follows a strong gradient, from high alpine (Aoraki/Mount Cook, 3724 m) through to the inter-montane Mackenzie valley (approx. 500 m) in which Lake Pukaki terminates. Mean annual precipitation in the headwaters has been estimated at 15 000 mm, falling to approximately 700 mm near the terminus of the lake (Kerr et al., 2011).
Daily lake inflow records are available for Lake Pukaki from the 1920s to the present day, and were obtained from Meridian Energy (the current operator of the hydro-electricity scheme for this lake). Daily precipitation data were sourced from the Hermitage/Mt Cook weather station near the head of Lake Pukaki (altitude 730 m), and used to confirm that inflow events were associated with a substantial rainfall event.
Previous research (Kingston et al., 2014) has identified September-December as the time of year with strongest linkages to atmospheric circulation patterns in the broader New Zealand region, so inflow records from these months were the focus here. The top 10 daily inflow events for each month (i.e. the top 10 September events, the top 10 October events, and so on) were identified. A month-by-month approach was used to avoid complications associated with the seasonal cycle.
The atmospheric drivers of the high inflow events were examined in two stages. Firstly, daily geopotential height fields at the time of high inflow events were analysed, using NCEP/NCAR reanalysis data (Kalnay et al., 1996). Geopotential height fields at 1000 hPa were examined initially to determine the extent to which daily atmospheric circulation patterns in the New Zealand region corresponded to the monthly relationships identified previously (Kingston et al., 2014). Additionally, larger-scale atmospheric circulation features during high inflow events were examined via 700 hPa geopotential heights across the extratropical southern hemisphere, to determine the connection between New Zealand regional circulation and more general features of the southern hemisphere atmosphere. Following previous identification of these indices as good descriptors of Waitaki lake inflow, the MZ1 and MZ2 were calculated on a daily time-step (following the procedure outlined by Salinger and Mullan (1999) for calculation of these indices with monthly data) to investigate whether extreme inflow events corresponded with extreme daily values of these indices. These indices describe the pressure gradient anomaly between Gisborne and Hokitika (MZ1) and Gisborne and Invercargill (MZ2). The availability of data from these locations resulted in these indices being calculated for the period 1960 to 2010 – in consequence this formed the period of analysis.
The second stage of the analysis involved determining the backwards trajectory of the air parcel over the approximate location of Lake Pukaki at the time of the high inflow event, and so provides further detail of the processes involved in delivering precipitation to the Pukaki catchment above that provided by daily circulation anomalies. Back-trajectory analysis was performed using the NOAA HYSPLIT tool (Draxler and Rolph, 2014), using NCEP/NCAR reanalysis data. Owing to uncertainty as to the exact timing during the day of peak inflow and peak precipitation, a series of trajectories were plotted for each event to ensure that the atmospheric drivers were captured. For each event, trajectories were started at four hour intervals, going backwards from midnight at the end of the day of the inflow peak, with six trajectories calculated in total. Trajectory duration was 72 h; an iterative process was followed to determine the most informative duration.
1000 hPa geopotential height fields for the top four September inflow events (9 September 1969, 17 September 1970, 13 September 1991 and 24 September 2008).
700 hPa geopotential height fields for the top four September inflow events (dates are the same as Fig. 1).
72 h trajectories for the top four September inflow events (dates are the same as Fig. 1). The red line indicates the trajectory started at midnight at the end of the day of each event; the five further trajectories were started at four-hour intervals preceding this (with the earliest trajectory indicated by the yellow line). For each year, the upper section indicates the path of the trajectory and the lower section indicates the relative humidity (%) from the start of the trajectory.
Trajectories were examined at sea level (rather than at a higher level in
the atmosphere) because a critical process for extreme precipitation
occurrence in this location is orographic uplift of moist airmasses
originating from the Tasman Sea (i.e. immediately northwest of the
catchment). As a result of the altitude of the Southern Alps, the moisture
content of the air at sea level can be an important indicator of the
likelihood of precipitation occurrence at the Main Divide. As such, even
though representation of the Southern Alps is limited given the
2.5
Geopotential height fields and back-trajectories associated with extreme
high inflow events can be divided into two general categories: the top four
September events neatly illustrate this division (Figs. 1–3). The first
category involves a pattern of high geopotential height (indicating an
anticyclone) immediately to the east of New Zealand, and low geopotential
height (i.e. a depression) to the west, as exemplified by the 1969 and 1991
September events (Fig. 1). Such a situation typically results in a strong
pressure gradient along a SW-NE axis across New Zealand, and is reflected in
strongly positive MZ1 and MZ2 index values. Analysis of Southern
Hemisphere-scale geopotential height at 700 hPa indicates that the main band
of midlatitude westerlies lie some distance south of New Zealand during
these events (Fig. 2). Trajectory analysis provides further detail about
the nature of atmospheric circulation upwind of Pukaki during these extreme
high inflow events (Fig. 3). The trajectories come from an approximate due
north direction, up to a latitude of approximately 30–35
The second general pattern of atmospheric circulation associated with
extreme high inflow indicates a much stronger zonal circulation over New
Zealand (and the South Island in particular) than for the first pattern, and
is exemplified here by the September 1970 and 2008 events (Fig. 1). A
trough of low pressure over New Zealand is typically present, but not a
fully formed depression as evident for the first pattern. The 700 hPa
geopotential height fields show that the increased zonal circulation over
New Zealand occurs because the main band of westerly winds is relatively far
north (in comparison to the first group; Fig. 2). The strong pressure
gradient associated with these mid-latitude westerlies results in strongly
positive MZ1 and MZ2 index values. The trajectories for this second group
indicate that air comes initially from a more north-northwesterly direction
(i.e. across the Tasman Sea) in comparison to the first pattern (Fig. 3).
This flow direction occurs only for the earliest few trajectory runs, after
which a more westerly trajectory is followed. The westerly orientated
trajectories cover a relatively large distance over the 72 h period of
analysis in comparison to the earlier northwesterly trajectories (and the
northerly trajectories of the first pattern). The north-northwesterly
trajectories are typically very humid (relative humidity
Results indicate two general patterns of atmospheric circulation (and so precipitation delivery) associated with extreme high inflow events for Lake Pukaki. The first pattern involves a pair of relatively slow moving weather systems over New Zealand: a departing anticyclone to the east, and an arriving depression to the west. Trajectory analysis suggests that in this situation Lake Pukaki is under the influence of warm, moist air (i.e. the warm “conveyor belt”) in advance of an arriving cold front; conditions often associated with heavy rainfall over the Southern Alps (Brenstrum, 1998). The consistency of the location of the trajectory indicates a relatively slow moving weather system, further strengthening the likelihood of high precipitation. Trajectories for these events typically approach or reach saturation at least one time-step (i.e. six hours) upwind of Pukaki. This level of humidity indicates that rainfall is likely to already be occurring as the air mass reaches the west coast of New Zealand. This is important because it has been shown previously that the heaviest precipitation events over the Southern Alps generally occur when orographic effects enhance existing synoptic-originating precipitation (Purdy and Austin, 2003).
As with the first pattern, the second group of geopotential height and
trajectory patterns indicates a warm moist air arriving from the north
associated with extreme inflow, again with relative humidity close to or at
100 % prior to reaching Lake Pukaki. However, pattern two differs in that
the arrival of this warm moist air occurs in a much more dynamic westerly
circulation, with precipitation associated with fronts embedded in a trough
rather than fully formed low pressure system (as was the case for the first
group). In this instance, the passage of the front and resultant change in
air mass humidity are clearly visible: relative humidity changes from
Analysis of daily MZ1 and MZ2 index values indicates that strong positive
association between these indices and monthly-resolution Pukaki inflow
(identified previously by Kingston et al., 2014
Conversations with Erick Brenstrum, Peter Kreft and Neil Gordon (MetService,
New Zealand) helped greatly with interpretation of the results. Jane McMecking
was supported by a University of Otago Department of Geography
summer scholarship. The authors gratefully acknowledge the NOAA Air
Resources Laboratory (ARL) for the provision of the HYSPLIT transport and
dispersion model and READY website (